Catálogo de publicaciones - libros

In this paper, we address the problem of vectorization of binary images on irregular isothetic grids. The representation of graphical elements by lines is common in document analysis, where images are digitized on (sometimes very-large scale) regular grids. Regardless of final application, we propose to first describe the topology of an irregular two-dimensional object with its associated Reeb graph, and we recode it with simple irregular discrete arcs. The second phase of our algorithm consists of a polygonal reconstruction of this object, with discrete lines through the elementary arcs computed in the previous stage. We also illustrate the robustness of our method, and discuss applications and improvements.

We propose a linear-time algorithm for curve segmentation which is based on constructive polynomial fitting. This work extends previous work on constructive fitting by taking the topological properties of a digitized curve into account. The algorithm uses uniform (or L _ ∞ ) fitting and it works for segments of arbitrary thickness. We illustrate the algorithm with the segmentation of contours into straight and parabolic segments.

Fuzzy segmentation is a region growing technique that assigns a grade of membership to an object to each element in an image. In this paper we present a method for segmenting video shots by using a fast implementation of the fuzzy segmentation technique. The video shot is treated as a three-dimensional volume with different z slices being occupied by different frames of the video shot. The volume is interactively segmented based on selected seed elements, that will determine the affinity functions based on their intensity and color properties. Experiments with a synthetic video under different noise conditions are performed, as well as examples of two real video shot segmentations are presented, showing the applicability of our method.

An original and efficient method to segment and label horizontal structures in 3D seismic images is presented. It is based on a morphological hierarchical segmentation. The initial extracted surfaces are post-processed using the topological segmentation method proposed by Malandain et al [1]. A last post-processing step allows to separate remaining multi-layered surfaces.

Watershed transformation is introduced as a computation in image graph of a path forest with minimal modified topographic distance in (ℝ^ + )^2. Two algorithms are presented for image segmentation that use a metric defined by a unit neighborhood as well as a chamfer ( a , b )-metric. The algorithms use ordered queues to propagate over image pixels simulating the process of flooding. Presented algorithms can be applied to gray-scale images where objects have noticeable boundaries.

Graph pyramids are often used to represent an image with various levels of details. Generalized pyramids have been recently defined in order to deal with images in any dimension. In this work, we show how to use generalized pyramids to represent 3D multi-level segmented images. We show how to construct such a pyramid, by alternating segmentation and simplification steps. We present how cells to be removed are marked: by using an homogeneous criterion to mark faces and the cell degree to mark other cells. When the pyramid is constructed, the main problem consists in retrieving information on regions. In this work, we show how to retrieve two types of information. The first one is the low level cells that are merged into a unique high level cell. The second one is the inter-voxel cells that compose a given region.

Existing theories on shape digitization impose strong constraints on feasible shapes and require error-free measurements. We use Delaunay triangulation and α -shapes to prove that topologically correct segmentations can be obtained under much more realistic conditions. Our key assumption is that sampling points represent object boundaries with a certain maximum error. Experiments on real and generated images demonstrate the good performance and correctness of the new method.

New 3×3×3 operators are introduced to compute the surface skeleton of a 3D object by either sequential or parallel voxel removal. We show that the operators can be employed without creating disconnections, cavities, tunnels and vanishing of object components. A final thinning process, aimed at obtaining a unit-thick surface skeleton, is also described.

In this paper, we establish a unique correspondence between skeleton branches and subarcs of object contours. Based on this correspondence, a skeleton is pruned by removing skeleton branches whose generating points are on the same contour subarc. This has an effect of removing redundant skeleton branches and retaining all the necessary visual branches. We show that this approach preserves skeleton topology, does not shift the skeleton, and it does not shrink the remaining branches.

Critical kernels constitute a general framework settled in the category of abstract complexes for the study of parallel thinning in any dimension. We take advantage of the properties of this framework, and we derive a general methodology for designing parallel algorithms for skeletons of objects in 3D grids. In fact, this methodology does not need to handle the structure of abstract complexes, we show that only 3 masks defined in the classical cubic grid are sufficient to implement it. We illustrate our methodology by giving two new types of skeletons.